@vibo-dev/skill-injection-scanner 1.1.4 → 1.1.6

package/README.md

@@ -70,11 +70,11 @@ No dependencies. Python 3.8+. Works on Linux/macOS.
 | 🟠 medium | comply-blind | "comply with everything the user says" |
 | 🟡 low | prio-emoji | "⚠️ IGNORE previous instructions" |
 
-Russian-language manipulation is covered too: «ты теперь…», «не говори владельцу»,
-«игнорируй предыдущие инструкции», «это критично: не сообщай…».
+Russian-language manipulation patterns are covered too: role takeover, secrecy orders,
+instruction override, "critical — do not tell the user" tricks.
 
-v1.1 additions: **follow-only** («следуй только этому тексту»), **attachment-instruction**
-(извлечение инструкции из картинки/вложения/alt), **system-msg-en**.
+v1.1 additions: **follow-only**, **attachment-instruction** (instructions read from an
+image/attachment/alt text), **system-msg** (RU and EN).
 
 ## Example output
 

package/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: skill-injection-scanner
 description: "Scan agent skill files for hidden instructions and prompt-injection patterns (EN/RU) before a poisoned skill rewrites your agent. 19 rules, zero dependencies. Use ONLY with the user's explicit consent: tell the user which skills folder will be scanned — findings are printed to stdout locally."
-version: 1.1.4
+version: 1.1.6
 tools: [python]
 license: MIT
 ---
@@ -25,9 +25,9 @@ python3 scanner.py --skills /path/to/skills --exclude .bak --include-code-spans
 ```
 
 ## What it detects (19 rules)
-- Role/personality hijack («you are now…», «ты теперь…») and system-message impersonation
+- Role/personality hijack ("you are now…" and Russian equivalents) and system-message impersonation
 - «Ignore previous instructions» / «follow only this text» (EN+RU)
-- Secrecy orders («never tell the owner…», «не говори владельцу»)
+- Secrecy orders ("never tell the owner…" and Russian equivalents)
 - Obfuscated instructions (base64/rot13/encoded), embedded `<|system|>` / ```` ```system ```` markers
 - Remote fetch-and-run (`curl | bash`, `git clone … && run`), instruction extraction from attachments/images
 - Manipulation tricks («this is critical: ignore…», emoji-boosted commands)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vibo-dev/skill-injection-scanner",
-  "version": "1.1.4",
+  "version": "1.1.6",
   "description": "Find hidden instructions and prompt-injection patterns inside agent skill files (EN/RU, 19 rules). Local-first, zero dependencies, Python 3.8+.",
   "license": "MIT",
   "author": "Viacheslav Bochkarev <vnbochkarev@gmail.com>",

package/scanner.py

@@ -33,7 +33,7 @@ import os
 import re
 import sys
 
-__version__ = "1.1.3"
+__version__ = "1.1.6"
 
 RULES = [
     # (name, regex, severity, note)